This article introduces my new peak MLB projections for pitching prospects, culminating in a ranking of the top 25 pitching prospects. In addition, peak projections for every minor and major league pitcher under 29 are now available at the Prospects Live Patreon page. Later in the offseason, these projections will join my peak projections for hitters at, https://scoutthestatline.com/, a website I co-publish with Ross Jensen.
Building aging curves and major league equivalencies for projecting pitchers
Broadly, these projections take performances at different ages, in different leagues, with different difficulties and scoring environments (but not park factors, as this data is not as easily accessible), and converts them all to the same peak American League baseline. This makes it easier to compare prospects with other prospects, and with major leaguers. If you’re not further interested in methodology, feel free to skip ahead to the top 25 projected pitching prospects list now (if you haven’t already!).
The first step in generating these projections was to create major league equivalencies to measure the difficulty of different leagues. To do this, I looked at the difference in performance from one season to the next in different leagues, then weighted the difference by the lesser of the total batters faced figures of the two seasons to find the average weighted change in performance (I just focused on league transitions from 2021 to 2022, with the recent minor league restructuring, but the results were broadly similar when looking at 2018 to 2019). I then chained the performance differences together to convert all leagues to the same American League baseline (e.g., CPX to A to A+ to AA to AAA to NL to AL). I chose to look at changes in league transitions from one season to the next, as I have found selection bias plays too big a role when looking at changes within the same season (from one season to the next, there is still selection bias, but it is at least partially offset by aging growth). The following paragraphs elaborate on the selection bias issue (which is similar for the league equivalencies and aging curves development processes).
The next step was to create aging curves for pitchers. I used the delta method, following past research from Tom Tango and Mitchel Lichtman, with my own methodological twists (of course, I take full responsibility for any mistakes or poor decisions I made!). The delta method looks at the difference in pitcher performance at each age from one season to the next. It then looks at the average change from season one to season two, weighted by some measure of playing time (in this case, total batters faced). The average change at each age is then chained together to form an overall aging curve.
The tricky part with the delta method is accounting for selection bias. When looking at season pairs, luckier pitchers in season one will tend to get more chances in season two, while unluckier pitchers will tend to get fewer chances. This bias understates aging growth and overstates aging decline, as luckier players will regress more in season two. Indeed, without accounting for selection bias, pitchers appear to decline from the very moment they set foot in the major leagues.
My sample for the aging curves covers all major league pitcher seasons from 1977 (the start of the free agency era) to 2022. I tested many different flavors of the delta method. In the end, I settled on the methodology that I found the most easy to defend. To account for selection bias, I created a projection so that pitchers received at least the same amount of playing time in season two as they did in season one. If pitchers had more playing time in season two than in season one, I didn’t adjust their numbers at all. If they had less, I created a projection for the “missing” playing time. I used a simple Marcel projection, weighting the past three seasons 3/2/1, from most recent to least, adding 1,200 total batters faced of regression to the mean. I then averaged the projection with the actual performance, weighted by playing time. With the selection bias accounted for, I proceeded to use the delta method as usual, looking at the change in performance from season one to season two at each age. I weighted this change by the total batters faced in season one, as selection bias should be a less significant issue for season one playing time. I then chained the changes together at each age to form an overall aging curve. I ended up extrapolating the aging curve from 22 to 31 to ages outside that range, as sample size and selection bias become problematic at younger and older ages (particularly for older ages, most pitchers retire and the survivors tend to be super good, understating aging decline, while at younger ages, there is little data, as most of those players are still in the minor leagues).
One note on the simple Marcel projection to forecast the missing data: I initially assumed no aging growth, deviating from the original Marcel slightly. This process produced an initial aging curve. I then applied my initial aging curve to the Marcel projections for the “missing” or “incomplete” seasons and repeated the delta method. I repeated this process until my aging curve stabilized.
Finally, I compared my aging curves with previous prominent research to make sure the overall growth was plausible—it was broadly consistent. As the main use of this research is for projecting prospects, I am less concerned with peak age and decline than with the overall growth of a player during their career.
Aging curves for ERA, K%, and BB%
The aging curves for K%, BB%, and ERA for a 19-year-old league-average pitcher are shown below. A 19-year-old league-average pitcher would be a phenom prospect, think of Andrew Painter, whom Steamer projected to be around league average by the end of the 2022 season.
According to my aging curves, a 19-year-old league average arm projects to be one of the best in the league by the time he hits his peak. Note that the peak age I find is 29 for ERA and K% minus BB%, with K% peaking a bit earlier and BB% peaking a bit later. This is generally a few years later of a peak for ERA and K% relative to most previous research, and perhaps a bit earlier of a peak for BB% (although comparable overall growth). My aging curves match previous work in finding that most growth occurs at young ages, though, with pitching talent remaining relatively stable throughout a pitcher’s peak years.
The top 25 projected starting pitching prospects
Enough with methodology. What sort of projections does this process generate? How do they compare with conventional prospecting wisdom? The top 25 projected starting pitching prospects are shown in the table below. The table is sorted by peak projected K% minus BB%, although peak projected xFIP would also have been a reasonable way to sort the players. The projections apply the well-known Marcel approach to my league equivalencies and aging curves, weighting past performance in 2022, 2021, and 2020, respectively, three to two to one, and applying some light regression to the mean (200 total batters faced, I used a light figure because I am projecting component statistics, K%, BB%, and FB%, that are relatively reliable in small samples). xFIP is a measure of K%, BB%, and FB%, assuming a league average HR/FB rate. Typically, Marcel assumes prospects will be league average. My equivalencies and aging curves allow for differentiation when projecting prospects.
Even though the projections ignore scouting entirely, it is reassuring to see the top projected starting pitching prospects are generally considered to be some of the game’s best pitching prospects, for the most part (damn you, Shawn Dubin, whom I could have excluded for too many relief appearances, but ultimately decided to keep him in there because he has had a decent chunk of appearances as a starter, too).
The top five especially each have an elite reputation among scouts, while the only surprise in the top 10 is Arizona’s Yu-Min Lin, who was fantastic in his debut across the CPX and Class A, as one of the youngest players in the latter league. Lin’s stuff does not have the same reputation as the rest of the top 10, with his velocity reportedly in the upper 80s to lower 90s depending on the source. He is anyway young enough that some velocity growth is a reasonable expectation.
As a bonus, I included Taj Bradley, Joey Cantillo, and Gavin Williams, three acclaimed pitching prospects who were not elite by K%, BB%, or xFIP, but have done an incredible job preventing runs so far in their minor league careers. A projection only capturing past ERAs would have each of them in the 3.40 to 3.50 range (although an ERA projection should regress more heavily to the mean as it is a higher variance metric subject to more luck, but still, even with the “full,” typical 1,200 total batters faced of regression, they’d all be below the 3.70 ERA mark). It remains to be seen whether they can improve their peripherals to match their run-prevention abilities, or whether the reverse will be true.
Finally, I included Brayan Bello, as this process also produces peak projections for major leaguers and includes all minor league and major league performance data. The future indeed appears to be bright for him despite an up-and-down debut, as he has shown an exceptional ability to keep the ball on the ground (his projected fly ball rate, a key metric in xFIP, is remarkably low). Similarly, Hunter Brown and the enigmatic, always old-for-level Brandon Walter get a nice boost when looking at xFIP instead of simply K% minus BB%: both project for an elite sub-3.4 xFIP, with a good K% minus BB%, paired with a fantastic FB%.
Wrapping up
Subscribers can now find peak projections for every minor and major league pitcher 29 or under at our Patreon page. They shouldn’t be taken as gospel, as they ignore scouting and stuff, and different sampling and methodology choices for producing the aging curves and league difficulty translations impact the ultimate output (assumptions always do, this is unavoidable!), but they can nonetheless hopefully provide readers with a useful tool, making it easier to compare pitchers pitching in different leagues, at different ages, by converting them all to the same major league baseline. At the very least, I’ll be leaning on them heavily in my own analyses moving forward—I invite you to take advantage of this when playing in dynasty leagues with me!